Image processing method and device, training method of neural network, and storage medium

A training method of a neural network, comprising:

• training a discriminative network based on a generative network to be trained;
• training the generative network to be trained based on the discriminative network;
• and
• alternately executing the above training processes to obtain a generative network,
• wherein the generative network is configured to process an input image network to generate an output image, the generative network comprises a first sub-network and at least one second sub-network, the first sub-network is configured to process the input image to obtain a plurality of first feature images, the at least one second sub-network is configured to perform a branching process and a weight sharing process on the plurality of first feature images to obtain a plurality of second feature images, and the plurality of second feature images is processed to obtain the output image;
• the training the generative network to be trained based on the discriminative network, comprises:
• processing a first training input image by using the generative network to be trained to generate a first training output image;
• calculating a system loss value of the generative network to be trained through a system loss function based on the first training output image;
• and
• tuning parameters of the generative network to be trained based on the system loss value,
• wherein the generative network further comprises a third sub-network, the plurality of second feature images is processed to obtain a plurality of third feature images;
• the plurality of third feature images is processed to obtain a plurality of fourth feature images by using the third sub-network;
• and the plurality of fourth feature images is performed through a synthesis process to obtain the output image,
• wherein the third sub-network comprises a second backbone network, a fourth branch network, a fifth branch network, and a sixth branch network, the plurality of third feature images is processed to obtain a plurality of fifth feature images by using the second backbone network, the plurality of fifth feature images is divided into a fourth branch input, a fifth branch input, and a sixth branch input, the fourth branch input is processed by using the fourth branch network to obtain a fourth feature image corresponding to the fourth branch network, the fifth branch input is processed by using the fifth branch network to obtain a fourth feature image corresponding to the fifth branch network, and the sixth branch input is processed by using the sixth branch network to obtain a fourth feature image corresponding to the sixth branch network;
• and
• wherein the plurality of fourth feature images comprises the fourth feature image corresponding to the fourth branch network, the fourth feature image corresponding to the fifth branch network and the fourth feature image corresponding to the sixth branch network.

An image processing device, comprising:

• a memory, configured to store computer readable instructions non-transitorily;
• and
• a processor, configured to execute the computer readable instructions, wherein
• upon the computer readable instructions being executed by the processor, an image processing method is executed, and the image processing method comprises:
• obtaining an input image;
• and
• processing the input image by using a generative network to generate an output image,
• wherein the generative network comprises a first sub-network and at least one second sub-network, and
• the processing the input image by using the generative network to generate the output image comprises:
• processing the input image by using the first sub-network to obtain a plurality of first feature images;
• performing a branching process and a weight sharing process on the plurality of first feature images by using the at least one second sub-network to obtain a plurality of second feature images;
• and
• processing the plurality of second feature images to obtain the output image,
• wherein the generative network further comprises a third sub-network, and
• the processing the plurality of second feature images to obtain the output image, comprises:
• processing the plurality of second feature images to obtain a plurality of third feature images;
• processing the plurality of third feature images by using the third sub-network to obtain a plurality of fourth feature images;
• and
• performing a synthesis process on the plurality of fourth feature images to obtain the output image,
• wherein the third sub-network comprises a second backbone network, a fourth branch network, a fifth branch network, and a sixth branch network,
• the processing the plurality of third feature images by using the third sub-network to obtain the plurality of fourth feature images, comprises:
• processing the plurality of third feature images by using the second backbone network to obtain a plurality of fifth feature images;
• dividing the plurality of fifth feature images into a fourth branch input, a fifth branch input, and a sixth branch input;
• and
• processing the fourth branch input by using the fourth branch network to obtain a fourth feature image corresponding to the fourth branch network, processing the fifth branch input by using the fifth branch network to obtain a fourth feature image corresponding to the fifth branch network, and processing the sixth branch input by using the sixth branch network to obtain a fourth feature image corresponding to the sixth branch network;
• and
• the plurality of fourth feature images comprises the fourth feature image corresponding to the fourth branch network, the fourth feature image corresponding to the fifth branch network and the fourth feature image corresponding to the sixth branch network.